Partisan Behavior of the Oklahoma State Legislature

In this project, I sought to both measure the partisan behavior of the members of the 2022 Regular Session of the Oklahoma State Legislature, as well as explain some of this partisan behavior by analyzing the demographic composition of Oklahoma at the district level.

Score Generation

I used the PSCL W-Nominate Package to generate partisan scores for both the House of Representatives and the Senate. These partisan scores are shown on the leaflets below:

Interpretation of These Scores

W-Nominate Scores are quite straightforward - a score of 1.0 represents the most conservative voting behavior possible, and a score of -1.0 represents the most liberal voting behavior possible. As seen on these choropleths, Senators in Oklahoma tend to be less polarized than House Representatives.

Analyzing Legislator Districts

Now that we have an understanding of which districts are the most polarized, we can attempt to determine why that is. From here on out, only the House of Representatives will be analyzed because the spread of partisan scores in the Senate is too small to visualize an appreciable difference in most cases.

First, this plot shows density of the ratio of residents earning incomes below the poverty line plotted by legislator party. We can observe that republican legislators in the house tend to be elected at a higher rate in districts where a smaller portion of residents are below the poverty line, while democrats are elected more evenly across poor and rich districts.

In this graph, the density of the ratio of females to males in each house district is plotted by legislator party. Interestingly, the gender ratio is more normally distributed in districts held by republicans than in districts held by democrats. This difference is likely not due to an underlying relationship, but rather to other district factors independent of gender and party.

In this graph, post-secondary educational attainment is plotted against legislator party. In this graph, it is clear that districts represented by democrats have higher levels of aggregate post-secondary educational attainment.

Finally, I ran W-Nominate scores against all of these factors in a linear regression model. From this model, it is evident that the poverty ratio, gender ratio, and educational attainment ratio of a district play weak but statistically significant roles in the partisan behavior of the legislator elected to represent it. While the strength of the correlation coefficients are low - likely due to the limited range of the dependent variable (W-Nominate score ranging from -1.0 - 1.0) - the moderate coefficient of determination in this model indicates that a real relationship exists between these variables.

## 
## ===============================================
##                         Dependent variable:    
##                     ---------------------------
##                           conserve_house       
## -----------------------------------------------
## poverty_ratio                -0.097***         
##                               (0.013)          
##                                                
## education_ratio              -0.070***         
##                               (0.009)          
##                                                
## male_ratio                   -0.080**          
##                               (0.035)          
##                                                
## Constant                     6.973***          
##                               (1.812)          
##                                                
## -----------------------------------------------
## Observations                    101            
## R2                             0.422           
## Adjusted R2                    0.404           
## Residual Std. Error       0.491 (df = 97)      
## F Statistic           23.561*** (df = 3; 97)   
## ===============================================
## Note:               *p<0.1; **p<0.05; ***p<0.01

Conclusion

From these data, a few things are clear. First, Oklahoma is an overwhelmingly conservative state, with the senate being less so than the house of representatives. Second, the gender composition, concentration of poverty, and educational attainment of a district can be used to predict the partisan behavior of the legislator elected to represent it.

I believe both party organizations in Oklahoma can benefit from these data. Democrats should first try to make a big push in Senate Districts held by outgoing incumbent moderates. Second, they should look to primarily target highly educated districts going forward.

For Republicans, I believe they should look to capitalize on their relative normal distribution across all three variables analyzed here and attempt to run viable candidates in every district. It is likely that the most highly-educated districts will not flip for them, but the other demographic factors of the district in question should be appraised before deciding to not run a challenger in Democrat-controlled districts.